WB tournaments: time control and participants

[Gabor Szots]

Geschrieben von:/Posted by: Gabor Szots at 23 February 2000 14:07:27:
Hi, everybody,
If you look at my two WB tournaments (GS-1 and GS-2), you'll see how well Gromit 2.20 did. A program which has only been given due attention since its new version came out.
Why? I think the main reason is time control. We would all like to organize as many tournaments as we can. Also, during testing new engines more information can be obtained if we play more games. The easiest way to achieve this is playing blitz games.
However, in blitz the quality of the game suffers, and positionally good engines have a great disadvantage. Such is the case with Gromit.
There are human tournaments with blitz and rapid time controls, but highly regarded are those with classic time control. Why should the case be different with machines? The better the evaluation function, the more time the engine will need for computing it.
My conclusion is that blitz is mainly for checking whether a program's functions work correctly, while it's chess value can only be measured at longer time controls. That means the usefulness of blitz tournaments is at least limited.
Am I right? I would appreciate your comments.
Kind regards,
Gabor

[Volker Pittlik]

Geschrieben von:/Posted by: Volker Pittlik at 23 February 2000 21:52:00:
Als Antwort auf:/As an answer to: WB tournaments: time control and participants geschrieben von:/posted by: Gabor Szots at 23 February 2000 14:07:27:
Hi Gabor!
In general I totally agree with you. When I'm testing an engine firstly I
let it play at very short time control (game in one minute) against an
engine which works fine at this time control (Crafty is a good choice
also for this purpose).
Secondly I like to whatch some games at usual blitz time control (3+2,
game in 5 or something like this).
If there are any problems with time control this problems will occur in
this two first steps.
What I would like to add is: for me it is just fun to whatch games at this
time control (mostly much more interesting than television). Also you
will get an allmost reliable impression of the strength of the engine.
I have never seen an engine winning at blitz mostly and loosing everytime on slow
time controls.
Lastly I let it play at longer time control to find the range of it's
real strength. At the moment I like to play game in an hour for this
purpose. I think there is not a great difference in playing strength
compared to 40 moves in two hours.
Best regards
Volker

Hi, everybody,
If you look at my two WB tournaments (GS-1 and GS-2), you'll see how well Gromit 2.20 did. A program which has only been given due attention since its new version came out.
Why? I think the main reason is time control. We would all like to organize as many tournaments as we can. Also, during testing new engines more information can be obtained if we play more games. The easiest way to achieve this is playing blitz games.
However, in blitz the quality of the game suffers, and positionally good engines have a great disadvantage. Such is the case with Gromit.
There are human tournaments with blitz and rapid time controls, but highly regarded are those with classic time control. Why should the case be different with machines? The better the evaluation function, the more time the engine will need for computing it.
My conclusion is that blitz is mainly for checking whether a program's functions work correctly, while it's chess value can only be measured at longer time controls. That means the usefulness of blitz tournaments is at least limited.
Am I right? I would appreciate your comments.
Kind regards,
Gabor

[Gabor Szots]

Geschrieben von:/Posted by: Gabor Szots at 24 February 2000 11:00:09:
Als Antwort auf:/As an answer to: Re: WB tournaments: time control and participants geschrieben von:/posted by: Volker Pittlik at 23 February 2000 21:52:00:
Hi Volker (and all who is interested),
I was glad to see your answer, the more so that it answered a question I haven't even asked: How long the time control should be? Game in half an hour seemed to be a good choice since it allowed concluding a tournament in about one or two week's time. But I still thought it was too short. On the other hand, 40 moves in 40 minutes (which I had thought of as the best choice) often produces too long games (I myself hardly bear the tension waiting for the final result). The one-hour game seems to be a good compromise.
By the way, do you think it good practice to end a game by force when it seems to have been decided?
As for blitz games, I am a slow thinker, so following a blitz game is almost impossible for me. Game in 15 minutes, on the other hand, is quite comfortable.
To the participant side:
I think it is best to play swiss tournaments. This way you achieve three goals:
(1) You finish within tolerable time.
(2) You can have bigger fields. The field can be inhomogenous, after a few rounds the programs "find their place" anyway. However, it is advisable to use seeding for the first round.
(3) The rating of newcomers is also possible.
Still, current practice seems to be round robin.
Best regards,
Gabor

[Volker Pittlik]

Geschrieben von:/Posted by: Volker Pittlik at 24 February 2000 21:17:03:
Als Antwort auf:/As an answer to: Re: WB tournaments: time control and participants geschrieben von:/posted by: Gabor Szots at 24 February 2000 11:00:09:
Hello Gabor,

I was glad to see your answer, the more so that it answered a question I >haven't even asked: How long the
time control should be? Game in half an hour seemed to be a good choice since it allowed
concluding a tournament in about one or two week's time. But I still thought >it was too short.
On the other hand,
40 moves in 40 minutes (which I had thought of as the best choice) often
produces too long games (I myself hardly bear the tension waiting for the >final result).
The one-hour game seems to be a good compromise.
By the way, do you think it good practice to end a game by force when it >seems to have been decided?
As for blitz games, I am a slow thinker, so following a blitz game is almost >impossible for me.
Game in 15 minutes, on the other hand, is quite comfortable.
To the participant side:
I think it is best to play swiss tournaments. This way you achieve three >goals:
(1) You finish within tolerable time.
(2) You can have bigger fields. The field can be inhomogenous, after a few rounds the programs "find their place" anyway. However, it is advisable to use seeding for the first round.
(3) The rating of newcomers is also possible.
Still, current practice seems to be round robin.

I cannot decide this. Maybe half an hour is enough. I think on average
a game is over after 60-70 moves. Ten to fifteen moves are played from
the book within seconds. So the engines have to play 50-60 moves with
normal time settings. So there are ~ 30 seconds for each move to
"think" about. I believe you will not find many different results of
the calculation of the engines comparing after 30 seconds and three
minutes. But of course you will find them sometimes.
Yes and most of the engines are very skilled to avoid 3-times
repetitions. I have seen games with 100 and more moves with this settings
very often. What means: 8-12 hours for _one_ game.
I hope so, maybe I will reduce to 45 minutes.
Hard to decide. The position must be _definitely_ won or a draw. I have
seen a game between Crafty (with 5-men tablebases) and another strong
engine without TBs. Crafty indicated a draw for a long time, but the
engine without TBs blundered and so Crafty has won that game.
Same for me.
Good points.
Volker

[Gabor Szots]

Geschrieben von:/Posted by: Gabor Szots at 25 February 2000 08:44:16:
Als Antwort auf:/As an answer to: Re: WB tournaments: time control and participants geschrieben von:/posted by: Volker Pittlik at 24 February 2000 21:17:03: